Parallelize riscv-arch-test suite

[piotro888]

It is currently artificially limited in CI workflow by NPROC=1 because of Verilator concurrency issues.
File locking approach from #554 could be probably used to fix this issue.

It may be more convenient to detect and run complied tests from python code via pytest, like it is done with riscv-test suite.
Simpler solution would be to include verilator file locking protection in run_signature.py script that is currently called for each target in the generated Makefile.